Enhanced Customer Retention: Deep Learning- Based Churn Prediction for Telecom Industry

Abstract

The Telecommunications Industry (TCI) faces a significant problem with customer attrition since revenue generation depends on keeping current customers. A deep learning-based architecture makes more accurate predictions about customer attrition. This brings to light a lot of problems. Churn aids in prioritizing new features or services that have the best chance of increasing customer retention. This guarantees that resources are directed toward the areas that can reduce churn the most.In this research, a deep learning-based model for predicting customer attrition in the telecom sector is presented. In order to extract sequential patterns from consumer behaviour, the model uses a 1D Convolutional Neural Network (CNN). By identifying spatial links in customer data, a 2D CNN improves feature extraction.The telecom statistics available on the Kaggle website aid in the prediction of churn in the telecom sector. Class imbalance in datasets is addressed by using the SMOTE, SMOTEEN, and SMOTETomek approaches. The performance analysis assesses recall rate, accuracy, precision, and F1 score. This methodical approach improves prediction accuracy.

Introduction

The telecom sector is highly competitive, and customer churn (users leaving a service) poses serious financial risks. Retaining customers is more cost-effective than acquiring new ones. Therefore, predicting and preventing churn is a top priority.

Churn Prediction Approach

Telecom companies use customer data (usage, preferences, history) to detect early signs of dissatisfaction. Traditional methods fall short due to the complex nature of churn, but modern deep learning and machine learning models have shown improved results.

Related Work

Several studies have employed ML and DL models for churn prediction:

• BiLSTM, CNN, Random Forest, Naïve Bayes, and hybrid models have achieved accuracies from ~88% to 98%.

• Techniques like SMOTE, attention mechanisms, and genetic algorithms have helped overcome issues like class imbalance and feature importance.

Methodology

A deep learning-based churn prediction system was developed using a dataset of 3,333 customers, with 483 churned and 2,850 retained.

1. Data Preprocessing

• Handled missing values, normalized features, and converted categorical variables.

• Balancing Techniques Used:

  • SMOTE: Generates synthetic minority class samples.

  • SMOTETomek: Removes overlapping samples between classes.

  • SMOTEENN: Combines oversampling and noise removal.

2. Model Training

Three deep learning models were trained:

A. 1D CNN (9 layers)
Designed for sequential data. Detects usage patterns.
Includes Conv1D, MaxPooling, Dense, Dropout, and sigmoid output.

B. 2D CNN (6 layers)
Applies filters in 2D to learn feature relationships.
Includes Conv2D, MaxPooling, Dense, Dropout, and sigmoid output.

C. Feedforward Neural Network (FNN)
Processes tabular customer data.
Includes multiple Dense layers with Batch Normalization and Dropout.

Model Evaluation Metrics

• Accuracy: Correct predictions (positive and negative) over total predictions.

• Precision: True positives among predicted positives.

• Recall: True positives among actual positives.

• F1-Score: Harmonic mean of precision and recall.

Key Takeaways

• Deep learning models, especially CNNs and FNNs, offer accurate churn prediction.

• Preprocessing and data balancing (SMOTE variants) are crucial for performance.

• Using these models, telecom companies can proactively reduce churn and improve customer loyalty.

Conclusion

The purpose of this paper was to use deep learning to build a reliable system for predicting customer churn in the telecom sector. To solve the class imbalance problem, three models FNN, 1D CNN, and 2D CNN were tested along with balancing techniques like SMOTE, SMOTETomek, and SMOTEENN. Each model was evaluated using accuracy, precision, recall, F1-score, and confusion matrices. Among them, the FNN model combined with SMOTEENN gave the best results, with accuracy improving from 88.6% to 96% and the F1-score increasing from 64 to 94. This shows that even a simple but well-tuned model can outperform more complex ones when working with structured data.

References

[1] N. Jajam, N. P. Challa, K. S. L. Prasanna, and C. H. V. S. Deepthi, “Arithmetic Optimisation With Ensemble Deep Learning SBLSTM-RNN-IGSA Model for Customer Churn Prediction,” IEEE Access, vol. 11, pp. 93111–93128, 2023, doi: 10.1109/ACCESS.2023.3304669. [2] I. Ullah, B. Raza, A. K. Malik, M. Imran, S. U. Islam, and S. W. Kim, “A Churn Prediction Model Using Random Forest: Analysis of Machine Learning Techniques for Churn Prediction and Factor Identification in Telecom Sector,” IEEE Access, vol. 7, pp. 60134–60149, 2019, doi: 10.1109/ACCESS.2019.2914999. [3] S. F. Bilal, A. A. Almazroi, S. Bashir, F. H. Khan, and A. A. Almazroi, “An ensemble-based approach using a combination of clustering and classification algorithms to enhance customer churn prediction in the telecom industry,” PeerJ Comput. Sci., vol. 8, p. e854, Feb. 2022. [4] A. Khattak, Z. Mehak, H. Ahmad, M. U. Asghar, M. Z. Asghar, and A. Khan, “Customer churn prediction using composite deep learning technique,” Sci. Rep., vol. 13, no. 1, Oct. 2023, Art. no. 17294. [5] A. Amin, S. Anwar, A. Adnan, M. Nawaz, N. Howard, J. Qadir, A. Hawalah, and A. Hussain, “Comparing oversampling techniques to handle the class imbalance problem: A customer churn prediction case study,” IEEE Access, vol. 4, pp. 7940–7957, 2016. [6] Y. Liu, M. Shengdong, G. Jijian, and N. Nedjah, “Intelligent Prediction of Customer Churn with a Fused Attentional Deep Learning Model,” Mathematics, vol. 10, no. 24, Dec. 2022, doi: 10.3390/math10244733. [7] F. E. Usman-Hamza et al., “Intelligent Decision Forest Models for Customer Churn Prediction,” Applied Sciences (Switzerland), vol. 12, no. 16, Aug. 2022, doi: 10.3390/app12168270. [8] I. V. Pustokhina, D. A. Pustokhin, P. T. Nguyen, M. Elhoseny, and K. Shankar, “Multi-objective rain optimisation algorithm with WELM model for customer churn prediction in telecommunication sector,” Complex and Intelligent Systems, vol. 9, no. 4, pp. 3473–3485, Aug. 2023, doi: 10.1007/s40747-021-00353- 6. [9] A. Amin et al., “Comparing Oversampling Techniques to Handle the Class Imbalance Problem: A Customer Churn Prediction Case Study,” IEEE Access, vol. 4, pp. 7940–7957, 2016, doi:10.1109/ACCESS.2016.2619719. [10] M. Óskarsdóttir, C. Bravo, W. Verbeke, C. Sarraute, B. Baesens, and J. Vanthienen, “Social Network Analytics for Churn Prediction in Telco: Model Building, Evaluation and Network Architecture,” Jan. 2020, doi:10.1016/j.eswa.2017.05.02 [11] C. P. Wei and I. T. Chiu, “Turning telecommunications call details to churn prediction: A data mining approach,” Expert Syst. Appl., vol. 23, no. 2, pp. 103– 112, Aug. 2002 [12] S. Saleh and S. Saha, “Customer retention and churn prediction in the telecommunication industry: a case study on a Danish university,” SN Appl Sci, vol. 5, no. 7, Jul. 2023, doi: 10.1007/s42452-023-05389-6. [13] A. Keramati, R. Jafari-Marandi, M. Aliannejadi, I. Ahmadian, M. Mozaffari, and U. Abbasi, “Improved churn prediction in telecommunication industry using data mining techniques,” Applied Soft Computing, vol. 24, pp. 994–1012, 2014, 2014,doi:10.1016/j.asoc.2014.08.041 [14] N. Alboukaey, A. Joukhadar, and N. Ghneim, “Dynamic behavior-based churn prediction in mobile telecom,” Expert Syst. Appl., vol. 162, Dec. 2020, Art. no. 113779 [15] A. A. K. Ahmad, A. Jafar, and K. Aljoumaa, “Customer churn prediction in telecom using machine learning in big data platform,” J Big Data, vol. 6, no. 1, Dec. 2019, doi: 10.1186/s40537-019-0191-6. [16]

Copyright

Copyright © 2025 S Surekha, B Yamuna. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.